http://ieeexplore.ieee.org/document/7949028/
This is tensorflow implementation for Deep Convolutional Neural Network for Inverse Problems in Imaging, TIP (2017)
.
Whole codes are forked and modified from https://github.com/jakeret/tf_unet.
Before starting,
pip install pillow matplotlib scipy scikit-image h5py
To start training a model for FBPConvNet:
python main.py --lr=1e-4 --output_path='logs/' --train_path='train_data/*.mat' --test_path='test_data/*.mat' --features_root=32 --layers=5
To deploy trained model:
python main.py --lr=1e-4 --output_path='logs/' --train_path='train_data/*.mat' --test_path='test_data/*.mat' --features_root=32 --layers=5 --is_training=False
You may find more details in main.py.
kyonghwan.jin@gmail.com